--> Inversion Case Studies From the SCOOP and STACK Areas in the Anadarko Basin
[First Hit]

2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Previous HitInversionNext Hit Case Studies From the SCOOP and STACK Areas in the Anadarko Basin

Abstract

The Mississippian section, and in particular the Meramec and the Devonian Woodford continue to be the preferred investment targets in the SCOOP/STACK trend in Oklahoma We showcase here the seismic characterization of these formations using multicomponent seismic Previous HitdataNext Hit in the STACK area and the conventional vertical component seismic Previous HitdataNext Hit in the SCOOP area, using deterministic prestack impedance Previous HitinversionNext Hit. The joint impedance Previous HitinversionNext Hit carried out over seismic Previous HitdataNext Hit from the STACK area was used to derive rock-physics parameters (Young’s modulus and Poisson’s ratio), which showed the sweet spots that are distinct spatially, rather than bleeding off at the edges. The added advantage of joint Previous HitinversionNext Hit was that the density attribute could also be derived therefrom, which was not possible for the Previous HitdataNext Hit from the STACK area. In addition to density, the results from prestack joint impedance Previous HitinversionNext Hit have been found to be superior to the simultaneous Previous HitinversionNext Hit. The equivalent attributes (besides density) derived for the SCOOP area also show promise. With the application of some of the unsupervised machine learning Previous HitmethodsNext Hit to Previous HitdataNext Hit from both areas, seismic facies classification was carried out for both the Meramec and Woodford intervals. Specifically, the principal component analysis (PCA), independent component analysis (ICA), kmeans clustering, self-organizing mapping (SOM) and generative topographic mapping (GTM) were applied to a suite of attributes derived from impedance Previous HitinversionNext Hit and by other means. These seismic facies were compared with similar facies derived for the intervals of interest from the well-Previous HitdataNext Hit, as well as other attributes. Within the unsupervised machine learning Previous HitmethodsTop, we found that ICA has an edge over PCA performance, and SOM and GTM provide additional information of interpretation interest. All these results will be demonstrated in our presentation.